Face identification with VGGFace and CelebA dataset

github.com/vlachvojta/POVa_face_identification

The goal of this project was to develop a facial recognition system from images. To achieve this, we im- plemented and evaluated a system utilizing two pre-trained models (VGGFace2 and ResNet50) alongside fine-tuning techniques with CelebA dataset. Our experiments demonstrated the effectiveness of attribute-focused training in im-proving recognition accuracy under challenging conditions such as sunglasses, lipstick, and blurry images. The results confirmed the robustness of the selected models and methodologies while highlighting areas for further optimization. The resulting demo application provides a practical example of face identification, showcasing its potential for user-friendly face recognition.